The VRN (Volumetric Regression Network) repository implements the “Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression” method. Instead of explicitly fitting a 3D model via landmark estimation and deformation, VRN treats the reconstruction task as volumetric segmentation: it learns a CNN to regress a 3D volume aligned to the input image, and then extracts a mesh via isosurface from that volume. The network is unguided (no 2D landmarks as intermediate). The mesh surfaces can be textured (in MATLAB branch) and colored. Docker container provided for easy CPU deployment.
Features
- Direct volumetric regression of 3D face (no landmarks)
- Mesh extraction from isosurface of predicted volume
- Optional texture / color mapping in MATLAB mode
- Docker container provided for easy CPU deployment
- GPU acceleration via CUDA / cuDNN
- Compatibility with Python / MATLAB based rendering / visualization
Categories
Computer Vision LibrariesLicense
MIT LicenseFollow VRN
Other Useful Business Software
Earn up to 16% annual interest with Nexo.
Generate interest, access liquidity without selling, and execute trades seamlessly. All in one platform.
Geographic restrictions, eligibility, and terms apply.
Rate This Project
Login To Rate This Project
User Reviews
Be the first to post a review of VRN!